blob: 13b8169793c568cfa2772fd636cbfc23ba7e3b01
1 | /* |
2 | * linux/init/main.c |
3 | * |
4 | * Copyright (C) 1991, 1992 Linus Torvalds |
5 | * |
6 | * GK 2/5/95 - Changed to support mounting root fs via NFS |
7 | * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96 |
8 | * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96 |
9 | * Simplified starting of init: Michael A. Griffith <grif@acm.org> |
10 | */ |
11 | |
12 | #define DEBUG /* Enable initcall_debug */ |
13 | |
14 | #include <linux/types.h> |
15 | #include <linux/module.h> |
16 | #include <linux/proc_fs.h> |
17 | #include <linux/kernel.h> |
18 | #include <linux/syscalls.h> |
19 | #include <linux/stackprotector.h> |
20 | #include <linux/string.h> |
21 | #include <linux/ctype.h> |
22 | #include <linux/delay.h> |
23 | #include <linux/ioport.h> |
24 | #include <linux/init.h> |
25 | #include <linux/initrd.h> |
26 | #include <linux/bootmem.h> |
27 | #include <linux/acpi.h> |
28 | #include <linux/tty.h> |
29 | #include <linux/percpu.h> |
30 | #include <linux/kmod.h> |
31 | #include <linux/vmalloc.h> |
32 | #include <linux/kernel_stat.h> |
33 | #include <linux/start_kernel.h> |
34 | #include <linux/security.h> |
35 | #include <linux/smp.h> |
36 | #include <linux/profile.h> |
37 | #include <linux/rcupdate.h> |
38 | #include <linux/moduleparam.h> |
39 | #include <linux/kallsyms.h> |
40 | #include <linux/writeback.h> |
41 | #include <linux/cpu.h> |
42 | #include <linux/cpuset.h> |
43 | #include <linux/cgroup.h> |
44 | #include <linux/efi.h> |
45 | #include <linux/tick.h> |
46 | #include <linux/interrupt.h> |
47 | #include <linux/taskstats_kern.h> |
48 | #include <linux/delayacct.h> |
49 | #include <linux/unistd.h> |
50 | #include <linux/rmap.h> |
51 | #include <linux/mempolicy.h> |
52 | #include <linux/key.h> |
53 | #include <linux/buffer_head.h> |
54 | #include <linux/page_ext.h> |
55 | #include <linux/debug_locks.h> |
56 | #include <linux/debugobjects.h> |
57 | #include <linux/lockdep.h> |
58 | #include <linux/kmemleak.h> |
59 | #include <linux/pid_namespace.h> |
60 | #include <linux/device.h> |
61 | #include <linux/kthread.h> |
62 | #include <linux/sched.h> |
63 | #include <linux/signal.h> |
64 | #include <linux/idr.h> |
65 | #include <linux/kgdb.h> |
66 | #include <linux/ftrace.h> |
67 | #include <linux/async.h> |
68 | #include <linux/kmemcheck.h> |
69 | #include <linux/sfi.h> |
70 | #include <linux/shmem_fs.h> |
71 | #include <linux/slab.h> |
72 | #include <linux/perf_event.h> |
73 | #include <linux/ptrace.h> |
74 | #include <linux/blkdev.h> |
75 | #include <linux/elevator.h> |
76 | #include <linux/sched_clock.h> |
77 | #include <linux/context_tracking.h> |
78 | #include <linux/random.h> |
79 | #include <linux/list.h> |
80 | #include <linux/integrity.h> |
81 | #include <linux/proc_ns.h> |
82 | #include <linux/io.h> |
83 | #include <linux/kaiser.h> |
84 | #include <linux/cache.h> |
85 | #ifdef CONFIG_AMLOGIC_PAGE_TRACE |
86 | #include <linux/amlogic/page_trace.h> |
87 | #endif |
88 | |
89 | #include <asm/io.h> |
90 | #include <asm/bugs.h> |
91 | #include <asm/setup.h> |
92 | #include <asm/sections.h> |
93 | #include <asm/cacheflush.h> |
94 | |
95 | static int kernel_init(void *); |
96 | |
97 | extern void init_IRQ(void); |
98 | extern void fork_init(void); |
99 | extern void radix_tree_init(void); |
100 | |
101 | /* |
102 | * Debug helper: via this flag we know that we are in 'early bootup code' |
103 | * where only the boot processor is running with IRQ disabled. This means |
104 | * two things - IRQ must not be enabled before the flag is cleared and some |
105 | * operations which are not allowed with IRQ disabled are allowed while the |
106 | * flag is set. |
107 | */ |
108 | bool early_boot_irqs_disabled __read_mostly; |
109 | |
110 | enum system_states system_state __read_mostly; |
111 | EXPORT_SYMBOL(system_state); |
112 | |
113 | /* |
114 | * Boot command-line arguments |
115 | */ |
116 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT |
117 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT |
118 | |
119 | extern void time_init(void); |
120 | /* Default late time init is NULL. archs can override this later. */ |
121 | void (*__initdata late_time_init)(void); |
122 | |
123 | /* Untouched command line saved by arch-specific code. */ |
124 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; |
125 | /* Untouched saved command line (eg. for /proc) */ |
126 | char *saved_command_line; |
127 | /* Command line for parameter parsing */ |
128 | static char *static_command_line; |
129 | /* Command line for per-initcall parameter parsing */ |
130 | static char *initcall_command_line; |
131 | |
132 | static char *execute_command; |
133 | static char *ramdisk_execute_command; |
134 | |
135 | /* |
136 | * Used to generate warnings if static_key manipulation functions are used |
137 | * before jump_label_init is called. |
138 | */ |
139 | bool static_key_initialized __read_mostly; |
140 | EXPORT_SYMBOL_GPL(static_key_initialized); |
141 | |
142 | /* |
143 | * If set, this is an indication to the drivers that reset the underlying |
144 | * device before going ahead with the initialization otherwise driver might |
145 | * rely on the BIOS and skip the reset operation. |
146 | * |
147 | * This is useful if kernel is booting in an unreliable environment. |
148 | * For ex. kdump situation where previous kernel has crashed, BIOS has been |
149 | * skipped and devices will be in unknown state. |
150 | */ |
151 | unsigned int reset_devices; |
152 | EXPORT_SYMBOL(reset_devices); |
153 | |
154 | static int __init set_reset_devices(char *str) |
155 | { |
156 | reset_devices = 1; |
157 | return 1; |
158 | } |
159 | |
160 | __setup("reset_devices", set_reset_devices); |
161 | |
162 | static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
163 | const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; |
164 | static const char *panic_later, *panic_param; |
165 | |
166 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; |
167 | |
168 | static bool __init obsolete_checksetup(char *line) |
169 | { |
170 | const struct obs_kernel_param *p; |
171 | bool had_early_param = false; |
172 | |
173 | p = __setup_start; |
174 | do { |
175 | int n = strlen(p->str); |
176 | if (parameqn(line, p->str, n)) { |
177 | if (p->early) { |
178 | /* Already done in parse_early_param? |
179 | * (Needs exact match on param part). |
180 | * Keep iterating, as we can have early |
181 | * params and __setups of same names 8( */ |
182 | if (line[n] == '\0' || line[n] == '=') |
183 | had_early_param = true; |
184 | } else if (!p->setup_func) { |
185 | pr_warn("Parameter %s is obsolete, ignored\n", |
186 | p->str); |
187 | return true; |
188 | } else if (p->setup_func(line + n)) |
189 | return true; |
190 | } |
191 | p++; |
192 | } while (p < __setup_end); |
193 | |
194 | return had_early_param; |
195 | } |
196 | |
197 | /* |
198 | * This should be approx 2 Bo*oMips to start (note initial shift), and will |
199 | * still work even if initially too large, it will just take slightly longer |
200 | */ |
201 | unsigned long loops_per_jiffy = (1<<12); |
202 | EXPORT_SYMBOL(loops_per_jiffy); |
203 | |
204 | static int __init debug_kernel(char *str) |
205 | { |
206 | console_loglevel = CONSOLE_LOGLEVEL_DEBUG; |
207 | return 0; |
208 | } |
209 | |
210 | static int __init quiet_kernel(char *str) |
211 | { |
212 | console_loglevel = CONSOLE_LOGLEVEL_QUIET; |
213 | return 0; |
214 | } |
215 | |
216 | early_param("debug", debug_kernel); |
217 | early_param("quiet", quiet_kernel); |
218 | |
219 | static int __init loglevel(char *str) |
220 | { |
221 | int newlevel; |
222 | |
223 | /* |
224 | * Only update loglevel value when a correct setting was passed, |
225 | * to prevent blind crashes (when loglevel being set to 0) that |
226 | * are quite hard to debug |
227 | */ |
228 | if (get_option(&str, &newlevel)) { |
229 | console_loglevel = newlevel; |
230 | return 0; |
231 | } |
232 | |
233 | return -EINVAL; |
234 | } |
235 | |
236 | early_param("loglevel", loglevel); |
237 | |
238 | /* Change NUL term back to "=", to make "param" the whole string. */ |
239 | static int __init repair_env_string(char *param, char *val, |
240 | const char *unused, void *arg) |
241 | { |
242 | if (val) { |
243 | /* param=val or param="val"? */ |
244 | if (val == param+strlen(param)+1) |
245 | val[-1] = '='; |
246 | else if (val == param+strlen(param)+2) { |
247 | val[-2] = '='; |
248 | memmove(val-1, val, strlen(val)+1); |
249 | val--; |
250 | } else |
251 | BUG(); |
252 | } |
253 | return 0; |
254 | } |
255 | |
256 | /* Anything after -- gets handed straight to init. */ |
257 | static int __init set_init_arg(char *param, char *val, |
258 | const char *unused, void *arg) |
259 | { |
260 | unsigned int i; |
261 | |
262 | if (panic_later) |
263 | return 0; |
264 | |
265 | repair_env_string(param, val, unused, NULL); |
266 | |
267 | for (i = 0; argv_init[i]; i++) { |
268 | if (i == MAX_INIT_ARGS) { |
269 | panic_later = "init"; |
270 | panic_param = param; |
271 | return 0; |
272 | } |
273 | } |
274 | argv_init[i] = param; |
275 | return 0; |
276 | } |
277 | |
278 | /* |
279 | * Unknown boot options get handed to init, unless they look like |
280 | * unused parameters (modprobe will find them in /proc/cmdline). |
281 | */ |
282 | static int __init unknown_bootoption(char *param, char *val, |
283 | const char *unused, void *arg) |
284 | { |
285 | repair_env_string(param, val, unused, NULL); |
286 | |
287 | /* Handle obsolete-style parameters */ |
288 | if (obsolete_checksetup(param)) |
289 | return 0; |
290 | |
291 | /* Unused module parameter. */ |
292 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) |
293 | return 0; |
294 | |
295 | if (panic_later) |
296 | return 0; |
297 | |
298 | if (val) { |
299 | /* Environment option */ |
300 | unsigned int i; |
301 | for (i = 0; envp_init[i]; i++) { |
302 | if (i == MAX_INIT_ENVS) { |
303 | panic_later = "env"; |
304 | panic_param = param; |
305 | } |
306 | if (!strncmp(param, envp_init[i], val - param)) |
307 | break; |
308 | } |
309 | envp_init[i] = param; |
310 | } else { |
311 | /* Command line option */ |
312 | unsigned int i; |
313 | for (i = 0; argv_init[i]; i++) { |
314 | if (i == MAX_INIT_ARGS) { |
315 | panic_later = "init"; |
316 | panic_param = param; |
317 | } |
318 | } |
319 | argv_init[i] = param; |
320 | } |
321 | return 0; |
322 | } |
323 | |
324 | static int __init init_setup(char *str) |
325 | { |
326 | unsigned int i; |
327 | |
328 | execute_command = str; |
329 | /* |
330 | * In case LILO is going to boot us with default command line, |
331 | * it prepends "auto" before the whole cmdline which makes |
332 | * the shell think it should execute a script with such name. |
333 | * So we ignore all arguments entered _before_ init=... [MJ] |
334 | */ |
335 | for (i = 1; i < MAX_INIT_ARGS; i++) |
336 | argv_init[i] = NULL; |
337 | return 1; |
338 | } |
339 | __setup("init=", init_setup); |
340 | |
341 | static int __init rdinit_setup(char *str) |
342 | { |
343 | unsigned int i; |
344 | |
345 | ramdisk_execute_command = str; |
346 | /* See "auto" comment in init_setup */ |
347 | for (i = 1; i < MAX_INIT_ARGS; i++) |
348 | argv_init[i] = NULL; |
349 | return 1; |
350 | } |
351 | __setup("rdinit=", rdinit_setup); |
352 | |
353 | #ifndef CONFIG_SMP |
354 | static const unsigned int setup_max_cpus = NR_CPUS; |
355 | static inline void setup_nr_cpu_ids(void) { } |
356 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
357 | #endif |
358 | |
359 | /* |
360 | * We need to store the untouched command line for future reference. |
361 | * We also need to store the touched command line since the parameter |
362 | * parsing is performed in place, and we should allow a component to |
363 | * store reference of name/value for future reference. |
364 | */ |
365 | static void __init setup_command_line(char *command_line) |
366 | { |
367 | saved_command_line = |
368 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); |
369 | initcall_command_line = |
370 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); |
371 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); |
372 | strcpy(saved_command_line, boot_command_line); |
373 | strcpy(static_command_line, command_line); |
374 | } |
375 | |
376 | /* |
377 | * We need to finalize in a non-__init function or else race conditions |
378 | * between the root thread and the init thread may cause start_kernel to |
379 | * be reaped by free_initmem before the root thread has proceeded to |
380 | * cpu_idle. |
381 | * |
382 | * gcc-3.4 accidentally inlines this function, so use noinline. |
383 | */ |
384 | |
385 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
386 | |
387 | static noinline void __ref rest_init(void) |
388 | { |
389 | int pid; |
390 | |
391 | rcu_scheduler_starting(); |
392 | /* |
393 | * We need to spawn init first so that it obtains pid 1, however |
394 | * the init task will end up wanting to create kthreads, which, if |
395 | * we schedule it before we create kthreadd, will OOPS. |
396 | */ |
397 | kernel_thread(kernel_init, NULL, CLONE_FS); |
398 | numa_default_policy(); |
399 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
400 | rcu_read_lock(); |
401 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
402 | rcu_read_unlock(); |
403 | complete(&kthreadd_done); |
404 | |
405 | /* |
406 | * The boot idle thread must execute schedule() |
407 | * at least once to get things moving: |
408 | */ |
409 | init_idle_bootup_task(current); |
410 | schedule_preempt_disabled(); |
411 | /* Call into cpu_idle with preempt disabled */ |
412 | cpu_startup_entry(CPUHP_ONLINE); |
413 | } |
414 | |
415 | /* Check for early params. */ |
416 | static int __init do_early_param(char *param, char *val, |
417 | const char *unused, void *arg) |
418 | { |
419 | const struct obs_kernel_param *p; |
420 | |
421 | for (p = __setup_start; p < __setup_end; p++) { |
422 | if ((p->early && parameq(param, p->str)) || |
423 | (strcmp(param, "console") == 0 && |
424 | strcmp(p->str, "earlycon") == 0) |
425 | ) { |
426 | if (p->setup_func(val) != 0) |
427 | pr_warn("Malformed early option '%s'\n", param); |
428 | } |
429 | } |
430 | /* We accept everything at this stage. */ |
431 | return 0; |
432 | } |
433 | |
434 | void __init parse_early_options(char *cmdline) |
435 | { |
436 | parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, |
437 | do_early_param); |
438 | } |
439 | |
440 | /* Arch code calls this early on, or if not, just before other parsing. */ |
441 | void __init parse_early_param(void) |
442 | { |
443 | static int done __initdata; |
444 | static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; |
445 | |
446 | if (done) |
447 | return; |
448 | |
449 | /* All fall through to do_early_param. */ |
450 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
451 | parse_early_options(tmp_cmdline); |
452 | done = 1; |
453 | } |
454 | |
455 | void __init __weak smp_setup_processor_id(void) |
456 | { |
457 | } |
458 | |
459 | # if THREAD_SIZE >= PAGE_SIZE |
460 | void __init __weak thread_stack_cache_init(void) |
461 | { |
462 | } |
463 | #endif |
464 | |
465 | /* |
466 | * Set up kernel memory allocators |
467 | */ |
468 | static void __init mm_init(void) |
469 | { |
470 | /* |
471 | * page_ext requires contiguous pages, |
472 | * bigger than MAX_ORDER unless SPARSEMEM. |
473 | */ |
474 | page_ext_init_flatmem(); |
475 | mem_init(); |
476 | #ifdef CONFIG_AMLOGIC_PAGE_TRACE |
477 | /* allocate memory before first page allocated */ |
478 | page_trace_mem_init(); |
479 | #endif |
480 | kmem_cache_init(); |
481 | percpu_init_late(); |
482 | pgtable_init(); |
483 | vmalloc_init(); |
484 | ioremap_huge_init(); |
485 | kaiser_init(); |
486 | } |
487 | |
488 | asmlinkage __visible void __init start_kernel(void) |
489 | { |
490 | char *command_line; |
491 | char *after_dashes; |
492 | |
493 | set_task_stack_end_magic(&init_task); |
494 | smp_setup_processor_id(); |
495 | debug_objects_early_init(); |
496 | |
497 | /* |
498 | * Set up the the initial canary ASAP: |
499 | */ |
500 | boot_init_stack_canary(); |
501 | |
502 | cgroup_init_early(); |
503 | |
504 | local_irq_disable(); |
505 | early_boot_irqs_disabled = true; |
506 | |
507 | /* |
508 | * Interrupts are still disabled. Do necessary setups, then |
509 | * enable them |
510 | */ |
511 | boot_cpu_init(); |
512 | page_address_init(); |
513 | pr_notice("%s", linux_banner); |
514 | setup_arch(&command_line); |
515 | mm_init_cpumask(&init_mm); |
516 | setup_command_line(command_line); |
517 | setup_nr_cpu_ids(); |
518 | setup_per_cpu_areas(); |
519 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
520 | boot_cpu_hotplug_init(); |
521 | |
522 | build_all_zonelists(NULL, NULL); |
523 | page_alloc_init(); |
524 | |
525 | pr_notice("Kernel command line: %s\n", boot_command_line); |
526 | /* parameters may set static keys */ |
527 | jump_label_init(); |
528 | parse_early_param(); |
529 | after_dashes = parse_args("Booting kernel", |
530 | static_command_line, __start___param, |
531 | __stop___param - __start___param, |
532 | -1, -1, NULL, &unknown_bootoption); |
533 | if (!IS_ERR_OR_NULL(after_dashes)) |
534 | parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, |
535 | NULL, set_init_arg); |
536 | |
537 | /* |
538 | * These use large bootmem allocations and must precede |
539 | * kmem_cache_init() |
540 | */ |
541 | setup_log_buf(0); |
542 | pidhash_init(); |
543 | vfs_caches_init_early(); |
544 | sort_main_extable(); |
545 | trap_init(); |
546 | mm_init(); |
547 | |
548 | /* |
549 | * Set up the scheduler prior starting any interrupts (such as the |
550 | * timer interrupt). Full topology setup happens at smp_init() |
551 | * time - but meanwhile we still have a functioning scheduler. |
552 | */ |
553 | sched_init(); |
554 | /* |
555 | * Disable preemption - early bootup scheduling is extremely |
556 | * fragile until we cpu_idle() for the first time. |
557 | */ |
558 | preempt_disable(); |
559 | if (WARN(!irqs_disabled(), |
560 | "Interrupts were enabled *very* early, fixing it\n")) |
561 | local_irq_disable(); |
562 | idr_init_cache(); |
563 | |
564 | /* |
565 | * Allow workqueue creation and work item queueing/cancelling |
566 | * early. Work item execution depends on kthreads and starts after |
567 | * workqueue_init(). |
568 | */ |
569 | workqueue_init_early(); |
570 | |
571 | rcu_init(); |
572 | |
573 | /* trace_printk() and trace points may be used after this */ |
574 | trace_init(); |
575 | |
576 | context_tracking_init(); |
577 | radix_tree_init(); |
578 | /* init some links before init_ISA_irqs() */ |
579 | early_irq_init(); |
580 | init_IRQ(); |
581 | tick_init(); |
582 | rcu_init_nohz(); |
583 | init_timers(); |
584 | hrtimers_init(); |
585 | softirq_init(); |
586 | timekeeping_init(); |
587 | time_init(); |
588 | sched_clock_postinit(); |
589 | printk_nmi_init(); |
590 | perf_event_init(); |
591 | profile_init(); |
592 | call_function_init(); |
593 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); |
594 | early_boot_irqs_disabled = false; |
595 | local_irq_enable(); |
596 | |
597 | kmem_cache_init_late(); |
598 | |
599 | /* |
600 | * HACK ALERT! This is early. We're enabling the console before |
601 | * we've done PCI setups etc, and console_init() must be aware of |
602 | * this. But we do want output early, in case something goes wrong. |
603 | */ |
604 | console_init(); |
605 | if (panic_later) |
606 | panic("Too many boot %s vars at `%s'", panic_later, |
607 | panic_param); |
608 | |
609 | lockdep_info(); |
610 | |
611 | /* |
612 | * Need to run this when irqs are enabled, because it wants |
613 | * to self-test [hard/soft]-irqs on/off lock inversion bugs |
614 | * too: |
615 | */ |
616 | locking_selftest(); |
617 | |
618 | #ifdef CONFIG_BLK_DEV_INITRD |
619 | if (initrd_start && !initrd_below_start_ok && |
620 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
621 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", |
622 | page_to_pfn(virt_to_page((void *)initrd_start)), |
623 | min_low_pfn); |
624 | initrd_start = 0; |
625 | } |
626 | #endif |
627 | page_ext_init(); |
628 | debug_objects_mem_init(); |
629 | kmemleak_init(); |
630 | setup_per_cpu_pageset(); |
631 | numa_policy_init(); |
632 | if (late_time_init) |
633 | late_time_init(); |
634 | sched_clock_init(); |
635 | calibrate_delay(); |
636 | pidmap_init(); |
637 | anon_vma_init(); |
638 | acpi_early_init(); |
639 | #ifdef CONFIG_X86 |
640 | if (efi_enabled(EFI_RUNTIME_SERVICES)) |
641 | efi_enter_virtual_mode(); |
642 | #endif |
643 | #ifdef CONFIG_X86_ESPFIX64 |
644 | /* Should be run before the first non-init thread is created */ |
645 | init_espfix_bsp(); |
646 | #endif |
647 | thread_stack_cache_init(); |
648 | cred_init(); |
649 | fork_init(); |
650 | proc_caches_init(); |
651 | buffer_init(); |
652 | key_init(); |
653 | security_init(); |
654 | dbg_late_init(); |
655 | vfs_caches_init(); |
656 | pagecache_init(); |
657 | signals_init(); |
658 | proc_root_init(); |
659 | nsfs_init(); |
660 | cpuset_init(); |
661 | cgroup_init(); |
662 | taskstats_init_early(); |
663 | delayacct_init(); |
664 | |
665 | check_bugs(); |
666 | |
667 | acpi_subsystem_init(); |
668 | sfi_init_late(); |
669 | |
670 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { |
671 | efi_late_init(); |
672 | efi_free_boot_services(); |
673 | } |
674 | |
675 | ftrace_init(); |
676 | |
677 | /* Do the rest non-__init'ed, we're now alive */ |
678 | rest_init(); |
679 | } |
680 | |
681 | /* Call all constructor functions linked into the kernel. */ |
682 | static void __init do_ctors(void) |
683 | { |
684 | #ifdef CONFIG_CONSTRUCTORS |
685 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
686 | |
687 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
688 | (*fn)(); |
689 | #endif |
690 | } |
691 | |
692 | bool initcall_debug; |
693 | core_param(initcall_debug, initcall_debug, bool, 0644); |
694 | |
695 | #ifdef CONFIG_KALLSYMS |
696 | struct blacklist_entry { |
697 | struct list_head next; |
698 | char *buf; |
699 | }; |
700 | |
701 | static __initdata_or_module LIST_HEAD(blacklisted_initcalls); |
702 | |
703 | static int __init initcall_blacklist(char *str) |
704 | { |
705 | char *str_entry; |
706 | struct blacklist_entry *entry; |
707 | |
708 | /* str argument is a comma-separated list of functions */ |
709 | do { |
710 | str_entry = strsep(&str, ","); |
711 | if (str_entry) { |
712 | pr_debug("blacklisting initcall %s\n", str_entry); |
713 | entry = alloc_bootmem(sizeof(*entry)); |
714 | entry->buf = alloc_bootmem(strlen(str_entry) + 1); |
715 | strcpy(entry->buf, str_entry); |
716 | list_add(&entry->next, &blacklisted_initcalls); |
717 | } |
718 | } while (str_entry); |
719 | |
720 | return 0; |
721 | } |
722 | |
723 | static bool __init_or_module initcall_blacklisted(initcall_t fn) |
724 | { |
725 | struct blacklist_entry *entry; |
726 | char fn_name[KSYM_SYMBOL_LEN]; |
727 | unsigned long addr; |
728 | |
729 | if (list_empty(&blacklisted_initcalls)) |
730 | return false; |
731 | |
732 | addr = (unsigned long) dereference_function_descriptor(fn); |
733 | sprint_symbol_no_offset(fn_name, addr); |
734 | |
735 | /* |
736 | * fn will be "function_name [module_name]" where [module_name] is not |
737 | * displayed for built-in init functions. Strip off the [module_name]. |
738 | */ |
739 | strreplace(fn_name, ' ', '\0'); |
740 | |
741 | list_for_each_entry(entry, &blacklisted_initcalls, next) { |
742 | if (!strcmp(fn_name, entry->buf)) { |
743 | pr_debug("initcall %s blacklisted\n", fn_name); |
744 | return true; |
745 | } |
746 | } |
747 | |
748 | return false; |
749 | } |
750 | #else |
751 | static int __init initcall_blacklist(char *str) |
752 | { |
753 | pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); |
754 | return 0; |
755 | } |
756 | |
757 | static bool __init_or_module initcall_blacklisted(initcall_t fn) |
758 | { |
759 | return false; |
760 | } |
761 | #endif |
762 | __setup("initcall_blacklist=", initcall_blacklist); |
763 | |
764 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
765 | { |
766 | ktime_t calltime, delta, rettime; |
767 | unsigned long long duration; |
768 | int ret; |
769 | |
770 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
771 | calltime = ktime_get(); |
772 | ret = fn(); |
773 | rettime = ktime_get(); |
774 | delta = ktime_sub(rettime, calltime); |
775 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; |
776 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", |
777 | fn, ret, duration); |
778 | |
779 | return ret; |
780 | } |
781 | |
782 | int __init_or_module do_one_initcall(initcall_t fn) |
783 | { |
784 | int count = preempt_count(); |
785 | int ret; |
786 | char msgbuf[64]; |
787 | |
788 | if (initcall_blacklisted(fn)) |
789 | return -EPERM; |
790 | |
791 | if (initcall_debug) |
792 | ret = do_one_initcall_debug(fn); |
793 | else |
794 | ret = fn(); |
795 | |
796 | msgbuf[0] = 0; |
797 | |
798 | if (preempt_count() != count) { |
799 | sprintf(msgbuf, "preemption imbalance "); |
800 | preempt_count_set(count); |
801 | } |
802 | if (irqs_disabled()) { |
803 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
804 | local_irq_enable(); |
805 | } |
806 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); |
807 | |
808 | add_latent_entropy(); |
809 | return ret; |
810 | } |
811 | |
812 | |
813 | extern initcall_t __initcall_start[]; |
814 | extern initcall_t __initcall0_start[]; |
815 | extern initcall_t __initcall1_start[]; |
816 | extern initcall_t __initcall2_start[]; |
817 | extern initcall_t __initcall3_start[]; |
818 | extern initcall_t __initcall4_start[]; |
819 | extern initcall_t __initcall5_start[]; |
820 | extern initcall_t __initcall6_start[]; |
821 | extern initcall_t __initcall7_start[]; |
822 | extern initcall_t __initcall_end[]; |
823 | |
824 | static initcall_t *initcall_levels[] __initdata = { |
825 | __initcall0_start, |
826 | __initcall1_start, |
827 | __initcall2_start, |
828 | __initcall3_start, |
829 | __initcall4_start, |
830 | __initcall5_start, |
831 | __initcall6_start, |
832 | __initcall7_start, |
833 | __initcall_end, |
834 | }; |
835 | |
836 | /* Keep these in sync with initcalls in include/linux/init.h */ |
837 | static char *initcall_level_names[] __initdata = { |
838 | "early", |
839 | "core", |
840 | "postcore", |
841 | "arch", |
842 | "subsys", |
843 | "fs", |
844 | "device", |
845 | "late", |
846 | }; |
847 | |
848 | static void __init do_initcall_level(int level) |
849 | { |
850 | initcall_t *fn; |
851 | |
852 | strcpy(initcall_command_line, saved_command_line); |
853 | parse_args(initcall_level_names[level], |
854 | initcall_command_line, __start___param, |
855 | __stop___param - __start___param, |
856 | level, level, |
857 | NULL, &repair_env_string); |
858 | |
859 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) |
860 | do_one_initcall(*fn); |
861 | } |
862 | |
863 | static void __init do_initcalls(void) |
864 | { |
865 | int level; |
866 | |
867 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
868 | do_initcall_level(level); |
869 | } |
870 | |
871 | /* |
872 | * Ok, the machine is now initialized. None of the devices |
873 | * have been touched yet, but the CPU subsystem is up and |
874 | * running, and memory and process management works. |
875 | * |
876 | * Now we can finally start doing some real work.. |
877 | */ |
878 | static void __init do_basic_setup(void) |
879 | { |
880 | cpuset_init_smp(); |
881 | shmem_init(); |
882 | driver_init(); |
883 | init_irq_proc(); |
884 | do_ctors(); |
885 | usermodehelper_enable(); |
886 | do_initcalls(); |
887 | } |
888 | |
889 | static void __init do_pre_smp_initcalls(void) |
890 | { |
891 | initcall_t *fn; |
892 | |
893 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
894 | do_one_initcall(*fn); |
895 | } |
896 | |
897 | /* |
898 | * This function requests modules which should be loaded by default and is |
899 | * called twice right after initrd is mounted and right before init is |
900 | * exec'd. If such modules are on either initrd or rootfs, they will be |
901 | * loaded before control is passed to userland. |
902 | */ |
903 | void __init load_default_modules(void) |
904 | { |
905 | load_default_elevator_module(); |
906 | } |
907 | |
908 | static int run_init_process(const char *init_filename) |
909 | { |
910 | argv_init[0] = init_filename; |
911 | return do_execve(getname_kernel(init_filename), |
912 | (const char __user *const __user *)argv_init, |
913 | (const char __user *const __user *)envp_init); |
914 | } |
915 | |
916 | static int try_to_run_init_process(const char *init_filename) |
917 | { |
918 | int ret; |
919 | |
920 | ret = run_init_process(init_filename); |
921 | |
922 | if (ret && ret != -ENOENT) { |
923 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", |
924 | init_filename, ret); |
925 | } |
926 | |
927 | return ret; |
928 | } |
929 | |
930 | static noinline void __init kernel_init_freeable(void); |
931 | |
932 | #if defined(CONFIG_DEBUG_RODATA) || defined(CONFIG_DEBUG_SET_MODULE_RONX) |
933 | bool rodata_enabled __ro_after_init = true; |
934 | static int __init set_debug_rodata(char *str) |
935 | { |
936 | return strtobool(str, &rodata_enabled); |
937 | } |
938 | __setup("rodata=", set_debug_rodata); |
939 | #endif |
940 | |
941 | #ifdef CONFIG_DEBUG_RODATA |
942 | static void mark_readonly(void) |
943 | { |
944 | if (rodata_enabled) |
945 | mark_rodata_ro(); |
946 | else |
947 | pr_info("Kernel memory protection disabled.\n"); |
948 | } |
949 | #else |
950 | static inline void mark_readonly(void) |
951 | { |
952 | pr_warn("This architecture does not have kernel memory protection.\n"); |
953 | } |
954 | #endif |
955 | |
956 | static int __ref kernel_init(void *unused) |
957 | { |
958 | int ret; |
959 | |
960 | kernel_init_freeable(); |
961 | /* need to finish all async __init code before freeing the memory */ |
962 | async_synchronize_full(); |
963 | free_initmem(); |
964 | mark_readonly(); |
965 | system_state = SYSTEM_RUNNING; |
966 | numa_default_policy(); |
967 | |
968 | rcu_end_inkernel_boot(); |
969 | |
970 | if (ramdisk_execute_command) { |
971 | ret = run_init_process(ramdisk_execute_command); |
972 | if (!ret) |
973 | return 0; |
974 | pr_err("Failed to execute %s (error %d)\n", |
975 | ramdisk_execute_command, ret); |
976 | } |
977 | |
978 | /* |
979 | * We try each of these until one succeeds. |
980 | * |
981 | * The Bourne shell can be used instead of init if we are |
982 | * trying to recover a really broken machine. |
983 | */ |
984 | if (execute_command) { |
985 | ret = run_init_process(execute_command); |
986 | if (!ret) |
987 | return 0; |
988 | panic("Requested init %s failed (error %d).", |
989 | execute_command, ret); |
990 | } |
991 | if (!try_to_run_init_process("/sbin/init") || |
992 | !try_to_run_init_process("/etc/init") || |
993 | !try_to_run_init_process("/bin/init") || |
994 | !try_to_run_init_process("/bin/sh")) |
995 | return 0; |
996 | |
997 | panic("No working init found. Try passing init= option to kernel. " |
998 | "See Linux Documentation/init.txt for guidance."); |
999 | } |
1000 | |
1001 | static noinline void __init kernel_init_freeable(void) |
1002 | { |
1003 | /* |
1004 | * Wait until kthreadd is all set-up. |
1005 | */ |
1006 | wait_for_completion(&kthreadd_done); |
1007 | |
1008 | /* Now the scheduler is fully set up and can do blocking allocations */ |
1009 | gfp_allowed_mask = __GFP_BITS_MASK; |
1010 | |
1011 | /* |
1012 | * init can allocate pages on any node |
1013 | */ |
1014 | set_mems_allowed(node_states[N_MEMORY]); |
1015 | /* |
1016 | * init can run on any cpu. |
1017 | */ |
1018 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1019 | |
1020 | cad_pid = task_pid(current); |
1021 | |
1022 | smp_prepare_cpus(setup_max_cpus); |
1023 | |
1024 | workqueue_init(); |
1025 | |
1026 | do_pre_smp_initcalls(); |
1027 | lockup_detector_init(); |
1028 | |
1029 | smp_init(); |
1030 | sched_init_smp(); |
1031 | |
1032 | page_alloc_init_late(); |
1033 | |
1034 | do_basic_setup(); |
1035 | |
1036 | /* Open the /dev/console on the rootfs, this should never fail */ |
1037 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) |
1038 | pr_err("Warning: unable to open an initial console.\n"); |
1039 | |
1040 | (void) sys_dup(0); |
1041 | (void) sys_dup(0); |
1042 | /* |
1043 | * check if there is an early userspace init. If yes, let it do all |
1044 | * the work |
1045 | */ |
1046 | |
1047 | if (!ramdisk_execute_command) |
1048 | ramdisk_execute_command = "/init"; |
1049 | |
1050 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { |
1051 | ramdisk_execute_command = NULL; |
1052 | prepare_namespace(); |
1053 | } |
1054 | |
1055 | /* |
1056 | * Ok, we have completed the initial bootup, and |
1057 | * we're essentially up and running. Get rid of the |
1058 | * initmem segments and start the user-mode stuff.. |
1059 | * |
1060 | * rootfs is available now, try loading the public keys |
1061 | * and default modules |
1062 | */ |
1063 | |
1064 | integrity_load_keys(); |
1065 | load_default_modules(); |
1066 | } |
1067 |